Sialoside Specificity of the Siglec Family Assessed Using Novel Multivalent Probes: Identification of Potent Inhibitors of Myelin-Associated Glycoprotein*

-linked carbohydrate

However, sialoadhesin is abundantly expressed on the surface of bone marrow macrophages and mediates the binding of immature granulocytes and has also been implicated in interactions with T cells during immune responses (18). MAG (Siglec-4) has been suggested to be important in maintenance of the myelin sheath and to inhibit neurite outgrowth in postnatal neurons (19,20).
While there is ample evidence that the functions of siglecs are modulated by the interaction with sialic acid containing ligands, establishing detailed mechanisms remains a challenge due to the ubiquitous distribution and structural diversity of sialyloligosaccharides on glycoproteins and glycolipids (3,7,11,21). More approachable has been the analysis of the specificity of siglecs to the sialoside sequence itself. The crystal structure for the sialoadhesin Vset domain with bound 3'-sialyllactose (NeuAcα(2-3)Galβ(1-4)Glc), revealed that sialoside binding is stabilized predominately by contacts with the sialic acid, including a salt bridge between the C-2 carboxyl-group and a highly conserved Arg at residue 97 (22). Yet, numerous reports have documented that the siglecs differ in their specificity towards other elements of sialoside sequences (2,3). For example, CD22 exhibits high specificity for the Siaα(2-6)Gal linkage (23)(24)(25), while sialoadhesin and MAG preferentially bind sialosides with the Siaα(2-3)Gal linkage (23,26), and Siglecs-7 and -11 exhibit preferred binding to sialosides with the Siaα(2-8)Sia linkage (1,27,28). There are also clear differences in the ability of siglecs to recognize naturally occurring sialic acids. Human sialoadhesin binds N-acetylneuraminic acid (NeuAc) but not N-glycolylneuraminic acid (NeuGc), while human CD22 binds both and murine CD22 preferentially binds NeuGc. Addition of a 9-O-acetyl group or truncation of the sialic acid to the C-7 analog with periodate abolishes binding of several siglecs examined (27,(29)(30)(31)(32)(33).
To date, most of the information obtained on siglec specificity has been gleaned from a variety of different assay methods involving the binding of recombinant siglecs or siglec-Fc chimeras to enzymatically modified erythrocytes, various multivalent sialoside probes, or immobilized gangliosides (23,24,27,28,31,(34)(35)(36)(37)(38). By their very nature, such multivalent assays typically yield qualitative information on the relative binding affinities of the sialoside sequences being studied. Several reports have evaluated sialoside specificity using competitive inhibition with monomeric sialosides yielding more quantitative comparisons of sialoside specificity, but only a few (1-3) of the siglecs have been compared in any given study (25,26,28,39,40). As a result, while important and useful information has been obtained, current information on the sialoside specificity of siglecs is somewhat fragmentary and does not readily yield to direct comparisons between members of the family.
In this report we have systematically examined the specificity of siglecs representing ten of the eleven members of the human siglec family or their murine orthologs against 28 synthetic and structurally defined sialosides representing the major terminal sialoside sequences found on carbohydrate groups of mammalian glycoproteins and glycolipids. This has been accomplished through the use of a novel and versatile multivalent sialoside probe platform based on adsorption of biotinylated synthetic sialosides to a commercial streptavidin-alkaline phosphatase conjugate (SAAP). Such probes can be rapidly assessed for binding siglec-Fc chimeras adsorbed to plastic wells of a microtiter plate. Binding is reversible and can be inhibited by free monovalent

EXPERIMENTAL PROCEDURES
Materials: Streptavidin-alkaline phosphatase (SAAP) was from Sigma (St. Louis, MO, cat. No. S2890). All other commercial chemical reagents used in this study were of highest purity available. α 1 -acid glycoprotein (α 1 -AGP) was a gift from the late Karl Schmid.
Biacore Assays: Flow cells of a CM5 sensor chip, equilibrated in HBS buffer (10 mM HEPES pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.005% surfactant P20) were activated with protein A (300 µg/mL, 10 mM NaOAc pH 4.8), at a flow rate of 10 µL/min for 10 min using standard EDC/NHS coupling (Biacore). Typically 12000 RU of protein A was immobilized. Sialidase (Vibrio cholerae) treated hCD22-Fc chimera or human IgG (7 µg/mL, as control), was loaded over the protein A flow cells until response units of 5000 RU were immobilized. All analyses were performed at a flow rate of 5 µL/min using HBS as eluent. Injection times were 3 min followed by 1 min dissociation. No regeneration after each sample was necessary. For competitive inhibition studies, sialosides ranging from 5-5000 µM, were mixed with a fixed concentration of α 1 -AGP (10 or 50 µM). Each experiment was followed by an internal control of the specific sialoside mixed with asialo-α 1 -AGP and subtracted from the sample containing the α 1 -AGP.
Subtractions were also made with the reference cell containing immobilized human IgG.

Synthesis of Monomeric Sialosides of Glycoprotein and
Glycolipids. The goal of this study was to investigate the relative specificity of 10 of the 11 known siglecs towards sialoside sequences that commonly occur in the glycans of potential natural ligands. For this purpose, sialosides representing terminal sequences on N-linked and O-linked carbohydrate groups of glycoproteins and on glycolipids were synthesized using chemoenzymatic methods, largely as described previously (41)(42)(43). Briefly, neutral core structures of Galβ(1-4)Glc (lactose), Galβ(1-3)GlcNAc (lacto-N-biose I) and Galβ(1-4)GlcNAc (N-acetyllactosamine or LacNAc) were synthesized with a short alkyl aglycon terminated with an azido group, and were then sialylated enzymatically in α2-3 and/or α2-6 linkage to galactose. Sialosides containing both NeuAc and NeuGc were synthesized in parallel since NeuAc is the predominant sialic acid in humans and NeuGc is a major sialic acid in mice (23)( Table I). These compounds were used as monovalent sialosides with no further modification, or were further modified by reduction of the azido group to the amine and conjugation to N-hydroxy succinimido-activated biotin as described in Experimental Procedures (Table I) (49,51,52,54,55). Our initial attempts to use these conjugates in an ELISA-type assay to quantitatively measure probe binding to siglec-Fc chimeras in the presence and absence of free sialosides, however, proved to be highly variable between experiments and even within a single experiment.
As a simple alternative to the sialoside-polyacrylamide probes, we investigated the use of probes comprised of sialoside-biotin conjugates adsorbed to streptavidin-alkaline phosphatase (SAAP) based on our preliminary experiments indicating that such probes would bind to siglec-Fc chimeras bound to wells of a protein A coated microtiter plate. Taking advantage of the high affinity of biotin for streptavidin, probes are simply prepared by mixing a biotinylated sialoside with a solution of SAAP (Sigma Chemical Company, S2890). Mixing at a ratio of 4 moles of sialoside per mole of streptavidin (approximately 2 nmol sialoside / 10 µg SAAP) gives optimal binding of the probes to immobilized siglec-Fc chimeras, with no increase in probe binding observed at higher ratios of sialoside to SAAP (data not shown).
A representative example of optimization of probe binding (NeuGcα(2-6)Galβ(1-4)GlcNAc-SAAP (L)) to a siglec-Fc chimera (Siglec-10) is shown in Fig. 2A. Various concentrations of siglec-Fc were added to protein A coated wells of a microtiter plate, and the adsorbed siglec-Fc chimera was overlaid with serially diluted sialoside-SAAP probe over a range of 0.04 to 5 µg/well. After 30 minutes the plates were washed and bound probe was detected with p-nitrophenyl phosphate (pNPP). Binding of the streptavidin-alkaline phosphatase probe was stable between 5 and 10 washes with buffer (data not shown). Saturation of the wells with siglec-Fc chimera typically occurred at a range of 0.25-0.5 µg well. For all experiments shown, each batch of purified siglec-Fc chimera was similarly analyzed prior to use to select an amount sufficient to saturate the wells of the microtiter plate.
The kinetics of the binding of sialoside-SAAP probe L to Siglec-10 is shown in Fig. 2B.
Binding of the probe increased gradually up to 60 minutes and was maximal by 120 minutes. No further increase in probe binding was apparent over 24 hours (not shown). In the presence of excess (4 mM) monovalent sialoside 20, probe binding was reduced to background. Moreover, addition of 4 mM sialoside 20 at 30 min reduced probe binding to the same level as when it was added at 0-time, indicating that binding of the sialoside-SAAP probe is reversible.
Two commercial sources of streptavidin-alkaline phosphtase (SAAP) were evaluated in this assay system. The product obtained from Sigma reproducibly yielded sialoside-SAAP probes that behaved as described above, while a similar product from Pierce yielded probes that showed no binding to siglec-Fc chimeras in the ELISA assay format. Analysis of the two preparations by a Superdex 200 column revealed that the Sigma product was entirely excluded in the void volume, while the Pierce product was heterogeneous and largely included. Analysis by SDS gel electrophoresis showed that both were heterogeneous, but the Sigma preparation had a distinct band at H 230-250 kDa and gradient of larger conjugates, while the Pierce material had predominately lower molecular weight species of 60-250 kDa (data not shown).
Siglec Specificity Towards the Sialoside-SAAP Conjugates. A panel of sixteen sialoside-SAAP probes (A-P, Table I) was prepared from their corresponding biotinylated sialosides as described in Experimental Procedures. Each was evaluated for their ability to bind to eleven siglec-Fc chimeras representing ten of the eleven known human siglecs (2-3 and 5-10) or their murine orthologs (1, 2 and 4). Probes were tested in duplicates for each siglec in at least two separate experiments, and results were highly reproducible when done on different days and with different batches of probe and siglec-Fc chimera. Results in Fig. 3 show representative experiments for ten of the siglecs. Each probe bound to a different set of siglecs except for probes C and G, which bound to the same set of six siglecs. Conversely, each siglec exhibited a unique specificity for binding the 16 sialoside-SAAP probes examined. At the extremes, mCD22  Table I). Yet, of the seven siglecs that bound to probe H, only Siglecs-9 and -10 bound to probe O with the short spacer.
This result clearly demonstrates that the length of spacer arm can influence the interaction of a sialoside sequence with a given siglec in this assay.
To determine the effect of siglec-Fc density on the microtiter plate on probe binding, several siglecs were evaluated for their ability to bind probes at sub-saturating levels. This was performed as a titration experiment examining the ability of all 16 probes to bind to Siglecs-1, -7 and -10 to microtiter wells coated with 2-fold dilutions of the respective siglec-Fc chimera.
Results shown in Fig. 4 compare the sialoside-SAAP specificity of each siglec coated at saturating levels (0.5 µg/well) to the specificity seen when coated at the lowest level that Siglec Specificity Towards Monomeric Sialosides. Binding of multivalent carbohydrate ligands to carbohydrate binding proteins is dependent on the structure of the carbohydrate as well as the nature of the multivalent display afforded by the carrier. In order to examine the specificity of the siglecs for sialoside sequences in the absence of the variables of multivalent display and steric interactions of the carrier, we sought to compare the relative affinity of monovalent sialosides for each siglec using a competition assay. To accomplish this, a single 'high affinity' sialoside-SAAP probe was selected for each siglec, and this was then used to detect siglec in the presence of increasing concentrations (4 nM -10 mM) of each monovalent sialoside. A sub-saturating level of probe was used to increase the sensitivity of the assay so the observed IC 50 value would approximate the K d value of the inhibitor being evaluated. A comparison of the inhibition curves obtained for selected sialosides with mCD22, hCD22, hSiglec-7 and hSiglec-10 are shown in  (Table II). Page 20
In this report seven human and four murine siglecs have been compared for their relative affinity for 28 monovalent sialosides comprising most of the terminal sequences found in carbohydrate groups of glycoproteins and glycolipids. Relative affinities were obtained as IC 50 values by competitive inhibition of the binding of a multivalent sialoside-SAAP probe, which was used at sub-saturating levels (See Fig. 5) to maximize the sensitivity of the inhibition such that the IC 50 value obtained would approximate its K d . In this regard, it is notable that the IC 50 value observed for sialoadhesin and its reference compound 5 (NeuAcα  (56). Providing that other siglecs exhibit a similar temperature dependence on affinity for their ligands, the true K d values at physiologic temperature (37°C) may be somewhat higher than the IC 50 values obtained at room temperature in this report.
The results shown in Table II allow detailed quantitative comparison of the specificity of the siglec family towards naturally occurring sialosides, and evaluation of the importance of oligosaccharide sequence on binding affinity. Striking is the fact that sialic acid (NeuAcαOMe, 2) is a very low affinity ligand for all siglecs examined with IC 50 values ranging from 2.4 mM to >10 mM. All siglecs exhibited higher affinities towards one or more of the various natural sialoside sequences evaluated. While each siglec exhibited a unique specificity, the increase in affinity afforded by the penultimate oligosaccharide sequence for the highest affinity sialoside ranged from as little as 3 fold for sialoadhesin (mSiglec-1) to greater than 12,000 fold for MAG . With the exception of MAG, the affinity exhibited for each siglec towards its preferred sialoside ligand is still low, with IC 50 values ranging from 0.14 -3.8 mM.
Several well documented elements of siglec specificity detected in multivalent assays are reflected in the sialoside affinities by the quantitative data summarized in Table II. The high specificity of human and murine CD22 (Siglec-2) for sialosides with the Siaα(2-6)Galβ(1-4)Glc(NAc) sequence (23,25), reflects a difference in binding affinity for α2-6 and α2-3 sialosides of at least 60 fold. No other siglec exhibited a preference for these α2-6 sialosides.
Comparison of the affinity of MAG for disialyl-T antigen (18) and its biosynthetically related O-linked oligosaccharides is summarized in Fig. 8. Striking is the observation that the non-sialylated Core 1 structure, Galβ(1-3)GalNAcα-OThr, shows no inhibition at 10 mM, while addition of sialic acid to either the 6 position of GalNAc or the 3 position of Gal increases affinity 1000 fold or more, and both sialic acids increase affinity over 30,000 fold. These results suggest that there may be two independent sites for binding sialic acid on MAG as suggested previously by others from analysis of MAG binding to ganglioside ligands (26,64,65). This conclusion was also reached by Vinson et al. who mutated the conserved Arg118 essential for sialic acid binding in sialoadhesin (22,60,66) and found that MAG still mediated sialic acid dependent effects on neurite outgrowth (67).
The ganglioside GD1α (and GQ1bα) contains the same terminal sequence as di-sialyl-T antigen (18), and has been reported to bind to MAG as a high affinity ligand (30). Yet, the free oligosaccharide of GD1α was found to be only 3 fold more potent an inhibitor of MAG than α2-3sialyllactose (4) (26), while the disialyl-T antigen was a 1500 fold more potent inhibitor than α2-3sialyllactose in this study. The primary difference between the structure of the GD1α oligosaccharide and disialyl-T antigen (18) is that the terminal oligosaccharide, NeuAcα(2- is linked β(1-4) to lactosylceramide in the former case, and is α linked to Thr in the latter. These differences appear to be critically important. Indeed, comparison of the linear sequence NeuAcα(2-3)Galβ(1-3)GalNAc linked in either β linkage (26) , or α linkage (27) to an aromatic aglycon shows that the α linkage is favored, and comparison of the aromatic aglycon (27) to the threonine aglycon (8) favors threonine by another 30 fold (Table II). Thus, both the α anomeric linkage, and the threonine aglycon contribute to the unique high affinity of the sialylated-T antigens (8, 17, and 18).
The relevance of the high affinity of MAG for these monomeric O-linked sialosides to natural ligands of MAG in situ is not clear at the present time. Several reports provide compelling evidence for gangliosides GT1b (and GD1a) as the major neuronal sialoside ligands for MAG (65,(67)(68)(69)(70). MAG has been shown to interact with glycoprotein carbohydrates in a sialic acid dependent manner (71). However, while MAG has been shown to bind to the neuronal Nogo receptor with high affinity, and that this interaction comprises a major mode of MAG induced inhibition of neurite outgrowth, binding to the Nogo receptor appears to be sialic acid independent (72,73). Regardless of the ultimate biological significance of the high affinity interactions with the analogs of O-linked sialosides reported here, we anticipate that such compounds and their derivatives will prove to be important tools in the analysis of in vitro and in vivo sialic acid dependent functions of MAG.
Although analysis of siglec specificity using multivalent sialosides has revealed many important elements of specificity seen with monovalent sialosides, multivalent presentation of ligands can differentially influence the observed specificity in several distinct ways. Multivalent constructs can amplify small differences in binding affinity, or differentially alter binding affinity through steric interactions between the carrier and the ligand-binding site. Amplification of small differences in affinity is well documented for differential binding of influenza virus hemagglutinin to NeuAcα(2-6)Gal and NeuAcα(2-3)Gal linkages on cell surface glycoprotein ligands. While the intrinsic binding affinity of the human H3 hemagglutinin to these ligands is 2.5 mM and 3.7 mM, respectively (74,75), this difference is amplified over 1000 fold in multivalent binding assays (76,77), and is sufficient to exert pressure for selection of receptor variants in natural and model host species (78,79). Such amplification of small differences in affinity can readily explain the observed preference of sialoadhesin for multivalent displays of α(2-3) sialosides over α(2-6) sialosides, and for NeuAc over NeuGc despite only 2-3 fold differences in affinity for the corresponding monovalent sialosides.
How two ligands might be differentially impacted by physical properties imposed by the carrier used in the multivalent constructs is more difficult to assess. Such factors are clearly evident from the data reported here and elsewhere. While both Siglec-5 and Siglec-7 exhibit similar affinity for sialosides with the NeuAcα(2-3)Gal and NeuAcα(2-6)Gal linkages (Table   II), only sialoside-SAAP probes with NeuAcα(2-3)Gal linkage bound to Siglec-5, and only probes with the NeuAcα(2-6)Gal linkage bound to Siglec-7 (Fig. 3). Yet binding of the same sialosides attached to polyacrylamide showed equivalent binding of both linkages to Siglec-5 and Siglec-7 (34,49,55,58,59). Conversely, high affinity binding of SLe X -SAAP (P) to sialoadhesin (Figs. 3 and 4) was not seen with the corresponding SLe X -polyacrylamide probe by Brinkman et al. (49). There is also little correlation between the intrinsic affinity of the sialosides (e.g. IC 50 values in Table II) and the ability of the corresponding SAAP probe to bind to immobilized siglec chimeras (Fig. 3). Indeed, the SAAP probe D (NeuAcα The systematic analysis of siglec specificity towards an extensive panel of sialosides was made possible by the use of a novel and flexible multivalent sialoside probe design consisting of biotinylated sialosides adsorbed to commercial streptavidin-alkaline phosphatase. Of paramount importance was the ability of the probes to bind to siglec chimeras adsorbed to a microtiter plate, and yield well behaved binding isotherms with low backgrounds and high reproducibility. In addition to their utility in microtiter plate-based binding assays, we have also found that the probes labeled with FITC are also well suited for binding to cell surface siglecs in flow cytometry experiments (unpublished results). These probes can be simply and inexpensively prepared with synthetic biotinylated sialosides as described here. Many of the biotinylated sialosides used in this report are also available with a shorter spacer from a commercial source (Glycotech) and with the long spacer from the Consortium for Functional Glycomics (http://web.mit.edu/glycomics/consortium/).    (Table I) Table II).  (Table I)         were prepared as described in Experimental Procedures. 2-azidoethyl-glycosides (A-C, E-H, K, L and O) were hydrogenolyzed over Pd/C followed by Biotinylation with N-hydroxy-succinimide activated-Biotins. O-linked sialosides (D, I, J and M) were biotinylated as decribed in Figure 1 and Experimental Procedures. The α-2,8sialoside-probe (N) was prepared from commercial disialyllactose (Sigma, St. Louis, MO) and biotinylated with amino-biotin (II) by reductive amination using sodium cyanoborohydride followed by acetic anhydride. Compound P was purchased from Glycotech (Bethesda, MD).   Table 2. Relative inhibitory potency of monovalent sialosides towards Siglecs 1-10. Monovalent sialosides were evaluated for their inhibitory potency toward each Siglec as described in Figure 5, using a matched sialoside-SAAP probe indicated by the * symbol in Figure 3 (for hSiglec-6, probe J was used). The value for each sialoside was calculated using the software program GraphPad. A reference compound for each Siglec was selected as the most potent sialoside inhibitor of the four N-acyl-sialyllactosamine compounds (5,9,13,20). Relative inhibitory potency for each sialoside is calculated as a percentage of the reference compound (100%) according to the formula: (Value) = (IC 50 reference compound / IC 50 sialoside) x 100.